四足機器人跨越非預知地形的最佳路徑規劃

Abstract

本研究延續本實驗室（CIDM Lab.）所開發出的四足機器人路徑規劃的理論，進一步探討發展針對非預先確定的地形進行最佳路徑的規劃方法。主要目的為提供四足機器人在非平坦地形或有動態障礙物等路況路徑規劃之應用，使機器人在跨越更複雜更多變地形的控制路徑也能快速正確地規劃出來，以期順利地到達目標點，本研究的成果可提高機器人自主行走或跨越於具高低起伏地形的功效。 由於四足機器人在非預先確知地形中跨足步伐隨前置步伐與後續步伐的考量選擇而牽動影響，且常不限於單一組解，本研究則選用並改良PRM的路徑規劃方法進行機器人任意單足跨越障礙物的連續路徑計算與規劃；其中並引用代克思托演算法(Dijkstra's algorithm)在特徵空間中尋找全域路徑模型中之最短路徑，以期規劃出跨越非確知地形的機器人之最佳路徑。本論文研究首先採用平面二連桿來測試本研究路徑規劃理論的正確與實用性，進而應用到本實驗室開發的NCF4仿生四足機器人的路徑規劃問題方面。在路徑規劃的程式系統執行步驟方面首先進行工作空間與特徵空間的轉換，建立起三個維度的特徵空間用做機器人任意單足的步伐姿態的定義與最佳路徑的搜尋評選之用。應用於四足機器人NCF4的非連續型地形中路徑規劃時，本研究的系統功能成功地支援NCF4仿生四足機器人的重心座標移動軌跡的評選與建構，此功能發展應用明顯可在NCF4仿生四足機器人路徑的動態模擬中明顯地觀測到較穩定較平衡的成效。

關鍵字：四足機器人、非確知地形、路徑規劃、代克思托演算法、PRM

The study proposed a novel path planning method for multilink robot walking on an indeterminacy terrain. The proposed method is based on the methodology of River-stone terrain trajectory planning method (provided by CIDM Lab.) and recruit Probabilistic Roadmap Method(PRM),Dijkstra shortest path planning searching Method etc. to develop a Matlab based Optimal path planning system with GUI interface. First of all, the robot configuration on a current terrain will be transfered into two to three more dimensional configuration space with obstacle marked area. Then the PRM path searching processes (including learning and query stages etc.) is modified by introducing the Dijkstra method to develop an auto planning software for generating an optimal path in predefined “n” dimensional configuration space. The free or obstacle avoiding path planning of two and three DOF robots have been solved by propose method as examples for demonstrating its practical performance. The current method has also been applied to generate the three variables quadruped robot motion planning task. The referenced simulation data have aproved that the helpful contribution on behalf of the proposed method.

Keywords: Quadruped robot, Configuration space, Path planning, Dijkstra's algorithm, PRM